ESTIMATING PROFILES OF FOREST ROADS BY GENETIC ALGORITHMS

摘要

This study examines a method for designing the optimum profile of a forest road in a short matter of time. In this method, a genetic algorithm (GA) searches for points in forest roads where the gradient changes, and dynamic programing (DP) designs the longitudinal slope. This method can be applied to the ground heights of an actual forest road. The inverse of the total of the construction expenditures estimated by DP and the change in longitudinal slope is the fitness function of GA. Crossover and mutation are each repeated 20 times and 420 profiles are designed and assessed, all within about 10 minutes. A comparison of the profile designed by the GA with that of an actual forest road shows only 2 sections where the formation level differed substantially, and the heights at 3 gradient change points nearly corresponded completely with those of an actual forest road. Thus, we believe that the profile designed by the GA is best, or the one of the best. Using the GA reduced the calculation time to about 1/389 that of the conventional method, showing that the method which uses the GA is very useful in designing profiles.